Medical schools are missing the mark on artificial intelligence

sLike it or not, healthcare is undergoing a massive transformation driven by artificial intelligence. But medical schools have barely begun to teach AI and machine learning — creating knowledge gaps that could exacerbate the damage done. Faulty algorithms And siding Decision support systems.

said Erkin Uetlich, a machine learning researcher working towards his medical degree and Ph.D. at the University of Michigan. “Without being armed with this set of basic knowledge about how these things work, we would be at a huge disadvantage.”

In a recent commentary published in Medicine Cell ReportsAnd Ötleş and a group of doctors and educators from the University of Michigan called on medical educators to make AI less an afterthought and more of a core concept in undergraduate medical training. They stress the idea of ​​a spiral curriculum, where students learn key points about AI in medicine at first, and then come back to it again and again as they learn more specialized skills.


This won’t be easy to implement, said co-author and former dean of the Medical College of Michigan, Jim Woolscroft. Bureaucratic inertia prevents medical school curricula from developing rapidly, and faculty members themselves may not have the expertise to educate a new generation of physicians. In an interview with STAT, the student and the teacher explained how medical educators can start the process of renewing AI training.

What is the current state of medical education in the field of artificial intelligence?


Jim Woolscroft: Medical education programs do not fundamentally evolve at all. There are a few adjustments, but not the kind of seismic changes that need to happen.

Erkin Oetsch: It’s not really specialized at all in terms of AI and machine learning. What you’ll often see, for those interested, is that they’ll either take some time to do either a master’s degree, or what I do, which is a Ph.D./MD. Otherwise, people may be exposed by the research they can do as electives. As a student, you still need to go and direct your own studies and need to inform yourself.

Is this enough? Or should it be artificial intelligence Is it integrated into the general medical curriculum?

Erkin Ötleş Courtesy of Stephanie Utsch

outlch: Artificial intelligence and machine learning will be so pervasive in our daily practice that everyone will need to have at least a basic level of understanding in order to evaluate the tools they use. They don’t need to be experts and they don’t need to develop these things, but they should be able to say, “I don’t think this works very well,” and then they can call the developer and say, “I think we have a problem.” We need to start teaching people quickly, because we’re going to be behind the eight ball.

Woolscroft: Medical students don’t know this stuff, and they need to see it as basic as pharmacology and physiology. Indeed, machine learning algorithms and artificial intelligence in general are basically everywhere.

One of the real problems is that our faculty members are not aware of this. Pre-Covid, I gave a talk on machine learning, and people were like, “Why does this matter?” They didn’t even know that in the university hospital, eight programs at the time were running continuously in the background, monitoring physiological variables on their patients. Teaching staff do not have the experience to teach.

When AI he It is taught in medical schools, what kind of strategies are used?

outlch: There is a lot of focus on certain technologies or tools, and that just doesn’t have to be the case. There was a discussion about: All these technologies use a lot of Python programming, so we should teach medical students Python programming. You know, I like Python programming, but I don’t think all my classmates who graduated from medical school should know how to program in Python.

Woolscroft: Many of the examples that seem to work, like in radiology, are really context based. thats good. But what we need is for the students to be at a much older level who understand some of the basic questions that need to be asked. What is the database, what care was taken to ensure that the data that was used to build the algorithm was clean? What gold standards were used? All of these things are broadly applicable questions.

So what is the most successful framework for AI medical education?

outlch: We need to prioritize teaching the basic concepts of AI and machine learning because we have limited time. We need to know the most important things and use them as a basis. And once you have that foundation, you can continually refer to it over time and then develop it if you need to or link it to other concepts.

Jim Woolscroft Courtesy of the University of Michigan

Woolscroft: One of the things that hasn’t been done, as far as I know, is the concept of the spiral curriculum: you come back to it over and over again as students move into clinical areas. So when they’re in the radiology department, they can ask: So, what is the interpretation of this mammogram? Did it include women from Egypt, for example, who had much more inflammatory breast cancer? you did not. good. Well, here in Michigan, we have a lot of people from the Middle East. Will this be applicable or not? When they get into all these different things, they will have a basis that they can connect to these specific examples of filling in the flesh of those bones that have been laid.

wWhat are the biggest obstacles to implementing this kind of curriculum change?

Woolscroft: Essentially, most medical schools have not changed their departmental structure to reflect changes in the science behind the practice of medicine. We have this structure, this legacy that leads to a lot of lethargy because there are all kinds of things attached to that, primarily budget and personnel.

The other real problem is making decisions about the curriculum, because this cannot be a supplement. Students will not see it as valuable. It has to be integrated and that requires the faculty to change a lot of the basic things that they were doing.

wThese are some of the first steps that need to be taken to overcome these barriers, and who needs to lead the attack?

outlch: We need to be led by doctors. They are likely to be academic medical centers or places with associates in colleges of engineering, computer science departments, colleges of information, or educational health systems departments. You will have resources that can merge together quickly – that’s what we need, we need that speed now.

And what about medical schools that don’t have these kinds of resources?

outlch: We’re trying to drive this as a conversation. We think having scaffolding — focusing on this core knowledge and then continually coming back to it — is an important way to put these things in place. People may think we’re completely off base, but hopefully they’ll agree that it’s important that we stand up and move quickly in this area. As part of that, we hope we can all think about how we share resources. When we build a curriculum, we share it, when we build a tool, we share it, so that we don’t waste time re-creating and can just jump into teaching and learning.

Woolscroft: it will happen. I’m just worried that it will happen more quickly. This is not unlike other technological innovations through the decades, even centuries: as new technology is introduced, often from outside of biology, it is applied to biological problems, discipline arises, and departments are created. I think it’s important enough that this be put on a fast track rather than just allowed to evolve organically. This cannot happen, because the patients will die.

This story is part of a series looking at the use of artificial intelligence in healthcare and the practices of sharing and analyzing patient data. It is supported by funding from Gordon and Betty Moore Foundation.

Leave a Comment